1 CounterfactualMetricBundle

pose_generation.CounterfactualMetricBundle(
    rri=None,
    root_gain=None,
    root_pm_dist=None,
    log_error_gain=None,
    target_rri=None,
    target_root_gain=None,
    target_root_pm_dist=None,
    target_log_error_gain=None,
    scene_rri=None,
    scene_root_gain=None,
    scene_root_pm_dist=None,
    scene_log_error_gain=None,
    target_pm_dist_before=None,
    target_pm_dist_after=None,
    target_pm_acc_before=None,
    target_pm_comp_before=None,
    target_pm_acc_after=None,
    target_pm_comp_after=None,
    target_candidate_support=None,
    target_current_support=None,
    scene_pm_dist_before=None,
    scene_pm_dist_after=None,
    scene_pm_acc_before=None,
    scene_pm_comp_before=None,
    scene_pm_acc_after=None,
    scene_pm_comp_after=None,
)

Typed per-valid-candidate metrics emitted by rollout evaluators.

1.1 Methods

Name	Description
from_vectors	Build a typed metric bundle from legacy or test vector mappings.
validate	Normalize all present metric vectors and verify valid-candidate length.
as_vectors	Return present metrics as name-to-vector mapping for trace serialization.

1.1.1 from_vectors

pose_generation.CounterfactualMetricBundle.from_vectors(vectors)

Build a typed metric bundle from legacy or test vector mappings.

1.1.2 validate

pose_generation.CounterfactualMetricBundle.validate(num_valid, device, dtype)

Normalize all present metric vectors and verify valid-candidate length.

1.1.3 as_vectors

pose_generation.CounterfactualMetricBundle.as_vectors()

Return present metrics as name-to-vector mapping for trace serialization.